Two percentages can describe the same Up/Down event while coming from different processes. The useful first step is to make sure they refer to the same market, outcome and observation time.
A model estimate is a forecast of the specified event’s outcome under the model’s data and assumptions. For OUTCORA’s current research, the scope is five-minute BTC, ETH, SOL and XRP Up/Down markets. A forecast should be read with its event identifier, direction and timestamp. It is not a statement that the model is certain, and it is not itself a market price.
A market-implied probability is derived from an observed event-market price. It is a time-specific market signal, not an independently verified truth about the future. The displayed probability can differ from a price available to a participant because of the bid–ask spread and market depth. Polymarket describes how its displayed prices are calculated in its price-calculation documentation; that display rule should not be assumed to describe every venue or every data feed.
Suppose an illustrative model estimate for Up is 64.8% and an observed market figure for the same event is 61.2%. Their arithmetic difference is 3.6 percentage points. These are example values, not a current OUTCORA observation. The comparison only makes sense when the event, outcome direction, settlement definition and observation time match.
The 3.6-point difference is not a projected gain. Spread, fees, liquidity, changing prices, timing and model error all matter. A difference can narrow or reverse before an event settles.
A useful record identifies the market and direction, observation timestamp, model estimate, observed market figure, data status and relevant spread, depth and time remaining. It also separates the reference price used for analysis from the venue’s official settlement outcome. Those fields let readers inspect the comparison without implying that a transaction was available or executed.
